In [1]:
### Data Analaysis
import numpy as np
import pandas as pd
### Data Visualization
import matplotlib.pyplot as plt
import seaborn as sns
import plotly.graph_objs as go
import plotly.express as px
from plotly.subplots import make_subplots
from scipy.interpolate import make_interp_spline, BSpline
%matplotlib inline
In [2]:
import warnings
warnings.simplefilter(action='ignore', category=Warning)
pd.set_option('display.max_columns', None)
pd.options.plotting.backend = "plotly"
confirmed_cases_df = pd.read_csv("https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_confirmed_global.csv")
recovered_cases_df = pd.read_csv("https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_recovered_global.csv")
death_cases_df = pd.read_csv("https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_deaths_global.csv")
1. Confirmed Cases¶
In [3]:
confirmed_cases_df.head()
Out[3]:
In [4]:
confirmed_cases_df.info()
Number of confirmed cases by country¶
In [5]:
confirmed_country = confirmed_cases_df.groupby('Country/Region').sum().drop(['Lat','Long'],axis=1)
confirmed_country.head()
Out[5]:
In [6]:
confirmed_country = confirmed_country.transpose()
confirmed_country.index
Out[6]:
In [7]:
confirmed_country.index.dtype
Out[7]:
In [8]:
# Set index as DateTimeIndex
datetime_index = pd.DatetimeIndex(confirmed_country.index)
confirmed_country.set_index(datetime_index, inplace=True)
# Check out index
confirmed_country.index
Out[8]:
In [9]:
countries = ['US','Brazil','Russia','India','United Kingdom','Spain','Italy','France','Iran','China']
confirmed_country[countries].plot(labels=dict(index="Date", value="Reported Confirmed cases count"),
title="Reported Confirmed Cases Time Series")
Number of confirmed cases by country (y-axis logarithmic)¶
In [10]:
countries = ['US','Brazil','Russia','India','United Kingdom','Spain','Italy','France','Iran','China']
confirmed_country[countries].plot(labels=dict(index="Date", value="Reported Confirmed cases count"),
title="Reported Confirmed Cases Time Series", log_y=True)
Countries with confirmed cases¶
In [11]:
#Latest date
confirmed_country_temp = confirmed_country.iloc[confirmed_country.shape[0]-1].to_frame()
confirmed_country_temp.columns = ['count']
confirmed_country_temp = confirmed_country_temp.reset_index()
In [12]:
fig = px.choropleth(confirmed_country_temp,
locations="Country/Region",
locationmode='country names',
color="count",
hover_name="Country/Region",
range_color=[1,500000],
color_continuous_scale="peach",
title='Countries with Confirmed Cases')
fig.show()
European countries with confirmed cases¶
In [13]:
europe = list(['Austria','Belgium','Bulgaria','Croatia','Cyprus','Czechia','Denmark','Estonia','Finland','France','Germany','Greece','Hungary','Ireland',
'Italy', 'Latvia','Luxembourg','Lithuania','Malta','Norway','Netherlands','Poland','Portugal','Romania','Slovakia','Slovenia',
'Spain', 'Sweden', 'United Kingdom', 'Iceland', 'Russia', 'Switzerland', 'Serbia', 'Ukraine', 'Belarus',
'Albania', 'Bosnia and Herzegovina', 'Kosovo', 'Moldova', 'Montenegro', 'North Macedonia'])
europe_confirmed_country_temp = confirmed_country_temp[confirmed_country_temp["Country/Region"].isin(europe)]
europe_confirmed_country_temp.head()
Out[13]:
In [14]:
fig = px.choropleth(europe_confirmed_country_temp,
locations="Country/Region",
locationmode='country names',
color="count",
hover_name="Country/Region", range_color=[1,200000],
color_continuous_scale='portland',
title='European Countries with Confirmed Cases',
scope='europe', height=800)
fig.show()
Confirmed cases spread over Time¶
In [15]:
confirmed_country_time = confirmed_country.unstack()
confirmed_country_time = confirmed_country_time.reset_index()
confirmed_country_time.columns = ["country","date","confirmed"]
confirmed_country_time['date'] = pd.to_datetime(confirmed_country_time['date'])
confirmed_country_time['date'] = confirmed_country_time['date'].dt.strftime('%m/%d/%Y')
confirmed_country_time.head()
Out[15]:
In [16]:
fig = px.scatter_geo(confirmed_country_time, locations="country",
locationmode='country names',
color="confirmed", size='confirmed',
hover_name="country",
range_color= [0, 100000],
projection="natural earth",
animation_frame="date",
title='COVID-19: Confirmed cases Spread Over Time',
color_continuous_scale="portland")
fig.show()
Trend Comparison of different countries¶
In [17]:
temp = confirmed_cases_df.groupby('Country/Region').sum().drop(["Lat","Long"],axis =1) \
.sort_values(confirmed_cases_df.columns[-1], ascending= False)
threshold = 50
f = plt.figure(figsize=(15,12))
ax = f.add_subplot(111)
for i,country in enumerate(temp.index):
if i >= 9:
if country != "India" and country != "Japan" :
continue
days = 90
t = temp.loc[temp.index== country].values[0]
t = t[t>threshold][:days]
date = np.arange(0,len(t[:days]))
xnew = np.linspace(date.min(), date.max(), 30)
spl = make_interp_spline(date, t, k=1) # type: BSpline
power_smooth = spl(xnew)
if country != "India":
plt.plot(xnew,power_smooth,'-o',label = country,linewidth =3, markevery=[-1])
else:
marker_style = dict(linewidth=4, linestyle='-', marker='o',markersize=10, markerfacecolor='#ffffff')
plt.plot(date,t,"-.",label = country,**marker_style)
plt.tick_params(labelsize = 14)
plt.xticks(np.arange(0,days,7),[ "Day "+str(i) for i in range(days)][::7])
# Reference lines
x = np.arange(0,18)
y = 2**(x+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate("No. of cases doubles every day",(x[-2],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,int(days-22))
y = 2**(x/2+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate(".. every second day",(x[-3],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,int(days-4))
y = 2**(x/7+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate(".. every week",(x[-3],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,int(days-4))
y = 2**(x/30+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate(".. every month",(x[-3],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
# India is following trend similar to doulbe the cases in 4 days but it may increase the rate
x = np.arange(0,int(days-5))
y = 2**(x/4+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "Red")
plt.annotate(".. every 4 days",(x[-3],y[-1]),color="Red",xycoords="data",fontsize=14,alpha = 0.8)
# plot Params
plt.xlabel("Days",fontsize=17)
plt.ylabel("Number of Confirmed Cases",fontsize=17)
plt.title("Trend Comparison of Confirmed Cases for Different Countries",fontsize=22)
plt.legend(loc = "upper left")
plt.yscale("log")
plt.grid(which="both")
#plt.savefig('Trend Comparison with India (confirmed).png')
plt.show()
2. Death Cases¶
In [18]:
deaths_country = death_cases_df.groupby('Country/Region').sum().drop(['Lat','Long'],axis=1)
deaths_country = deaths_country.transpose()
# Set index as DateTimeIndex
datetime_index = pd.DatetimeIndex(deaths_country.index)
deaths_country.set_index(datetime_index, inplace=True)
deaths_country.head()
Out[18]:
Number of reported deaths by country¶
In [19]:
countries = ['US','Brazil','Russia','India','United Kingdom','Spain','Italy','France','Iran','China']
deaths_country[countries].plot(labels=dict(index="Date", value="Reported Death cases count"),
title="Reported Death Cases Time Series")
Number of reported deaths by country (y-axis logarithmic)¶
In [20]:
countries = ['US','Brazil','Russia','India','United Kingdom','Spain','Italy','France','Iran','China']
deaths_country[countries].plot(labels=dict(index="Date", value="Reported Death cases count"),
title="Reported Death Cases Time Series", log_y=True)
Death cases spread over Time¶
In [21]:
deaths_country_time = deaths_country.unstack()
deaths_country_time = deaths_country_time.reset_index()
deaths_country_time.columns = ["country","date","death"]
deaths_country_time['date'] = pd.to_datetime(deaths_country_time['date'])
deaths_country_time['date'] = deaths_country_time['date'].dt.strftime('%m/%d/%Y')
deaths_country_time.head()
Out[21]:
In [22]:
fig = px.scatter_geo(deaths_country_time, locations="country",
locationmode='country names',
color="death", size='death',
hover_name="country",
range_color= [0, 10000],
projection="natural earth",
animation_frame="date",
title='COVID-19: Death Cases Spread Over Time',
color_continuous_scale="portland")
fig.show()
3. Recovered Cases¶
In [23]:
recovered_country = recovered_cases_df.groupby('Country/Region').sum().drop(['Lat','Long'],axis=1)
recovered_country = recovered_country.transpose()
# Set index as DateTimeIndex
datetime_index = pd.DatetimeIndex(recovered_country.index)
recovered_country.set_index(datetime_index, inplace=True)
recovered_country.head()
Out[23]:
Number of recovered cases by country¶
In [24]:
countries = ['US','Brazil','Russia','India','United Kingdom','Spain','Italy','France','Iran','China']
recovered_country[countries].plot(labels=dict(index="Date", value="Recovered cases count"),
title="Recovered Cases Time Series")
Number of recovered cases by country (y-axis logarithmic)¶
In [25]:
countries = ['US','Brazil','Russia','India','United Kingdom','Spain','Italy','France','Iran','China']
recovered_country[countries].plot(labels=dict(index="Date", value="Recovered cases count"),
title="Recovered Cases Time Series", log_y=True)
Recovered cases spread over Time¶
In [26]:
recovered_country_time = recovered_country.unstack()
recovered_country_time = recovered_country_time.reset_index()
recovered_country_time.columns = ["country","date","recovered"]
recovered_country_time['date'] = pd.to_datetime(recovered_country_time['date'])
recovered_country_time['date'] = recovered_country_time['date'].dt.strftime('%m/%d/%Y')
recovered_country_time.head()
Out[26]:
In [27]:
fig = px.scatter_geo(recovered_country_time, locations="country",
locationmode='country names',
color="recovered", size='recovered',
hover_name="country",
range_color= [0, 10000],
projection="natural earth",
animation_frame="date",
title='COVID-19: Recoveries Spread Over Time',
color_continuous_scale="portland")
fig.show()
Comparison between Recovered, Deaths and Active cases¶
In [28]:
def remove_dup_columns(frame):
keep_names = set()
keep_icols = list()
for icol, name in enumerate(frame.columns):
if name not in keep_names:
keep_names.add(name)
keep_icols.append(icol)
return frame.iloc[:, keep_icols]
combined_country = pd.concat([confirmed_country_time, deaths_country_time, recovered_country_time],axis=1)[['country','date','confirmed','death','recovered']]
combined_country = remove_dup_columns(combined_country)
combined_country['active'] = combined_country['confirmed'] - combined_country['death'] - combined_country['recovered']
combined_country['mortality_rate'] = np.round((combined_country['death']/combined_country['confirmed'])*100,2)
combined_country.tail()
Out[28]:
In [29]:
x = combined_country['date']
y = combined_country['confirmed']
z = combined_country['death']
w = combined_country['recovered']
a = combined_country['active']
#call Bar charts and assign to fig_t. Add traces to the same figure
fig_t = go.Figure(go.Bar(x=x, y=y, name='Total Confirmed', marker_color='indianred', opacity=.7))
fig_t.add_trace(go.Bar(x=x, y=a, name='Total Active', marker_color='mediumblue', opacity=0.7))
fig_t.add_trace(go.Bar(x=x, y=w, name='Total recovered', marker_color='lightseagreen', opacity=0.7))
fig_t.add_trace(go.Bar(x=x, y=z, name='Total deaths', marker_color='gray', opacity=1))
#here we define layout of the chart
fig_t.update_layout(barmode='overlay',
xaxis={'categoryorder':'total ascending'},
xaxis_type='category',
title={
'text': 'Cummulative COVID-19 Trend',
'y':0.79,
'x':0.45,
'xanchor': 'center',
'yanchor': 'top'},)
fig_t.update_xaxes(title= '----->Timeline' ,showline=True)
fig_t.update_yaxes(title= '----->Number of cases', showline=True)
fig_t.show()
Mortality Rate¶
In [30]:
x = combined_country['date']
m = combined_country['mortality_rate']
fig_t = go.Figure(go.Bar(x=x, y=m, name='Mortality Rate', marker_color='indianred', opacity=.7))
#here we define layout of the chart
fig_t.update_layout(barmode='overlay',
xaxis={'categoryorder':'total ascending'},
xaxis_type='category',
title={
'text': 'Cummulative Mortality Rate Trend',
'y':0.79,
'x':0.45,
'xanchor': 'center',
'yanchor': 'top'},)
fig_t.update_xaxes(title= '----->Timeline' ,showline=True)
fig_t.update_yaxes(title= '----->Mortality Rate', showline=True)
fig_t.show()
Comparison between Active, Recovered, Confirmed and Death Cases for India¶
In [31]:
india_data = combined_country[combined_country["country"]=='India']
In [32]:
x = india_data['date']
y = india_data['confirmed']
z = india_data['death']
w = india_data['recovered']
a = india_data['active']
#call Bar charts and assign to fig_t. Add traces to the same figure
fig_t = go.Figure(go.Bar(x=x, y=y, name='Total Confirmed', marker_color='indianred', opacity=.7))
fig_t.add_trace(go.Bar(x=x, y=a, name='Total Active', marker_color='mediumblue', opacity=0.7))
fig_t.add_trace(go.Bar(x=x, y=w, name='Total recovered', marker_color='lightseagreen', opacity=0.7))
fig_t.add_trace(go.Bar(x=x, y=z, name='Total deaths', marker_color='gray', opacity=1))
#here we define layout of the chart
fig_t.update_layout(barmode='overlay',
xaxis={'categoryorder':'total ascending'},
xaxis_type='category',
title={
'text': 'Cummulative COVID-19 Trend of India',
'y':0.79,
'x':0.45,
'xanchor': 'center',
'yanchor': 'top'},)
fig_t.update_xaxes(title= '----->Timeline' ,showline=True)
fig_t.update_yaxes(title= '----->Number of cases', showline=True)
fig_t.show()
Time-Series Forecasting¶
In [33]:
confirmed_india = confirmed_country['India'].reset_index().rename(columns={'index':'ds','India':'y'})
confirmed_india['ds'] = pd.to_datetime(confirmed_india['ds'])
confirmed_india.head()
Out[33]:
Prophet is a procedure for forecasting time series data based on an additive model where non-linear trends are fit with yearly, weekly, and daily seasonality, plus holiday effects. It works best with time series that have strong seasonal effects and several seasons of historical data. Prophet is robust to missing data and shifts in the trend, and typically handles outliers well.
Prophet is open source software released by Facebook’s Core Data Science team.
In [34]:
from fbprophet import Prophet
In [35]:
model = Prophet(interval_width=0.95)
model.fit(confirmed_india)
future_7 = model.make_future_dataframe(periods=7)
future_7.tail()
Out[35]:
In [36]:
#predicting the future with date, upper and lower limit of y values
forecast = model.predict(future_7)
forecast[['ds','yhat','yhat_lower','yhat_upper']].tail()
Out[36]:
In [37]:
confirmed_forecast_plot = model.plot(forecast)
In [38]:
confirmed_forecast_plot = model.plot_components(forecast)
